Model Predictive Control (MPC) refers to a class of algorithms that compute a sequence of manipulated variable adjustments in order to optimize the future behavior of complex multivariable processes. Originally developed to meet the needs of petroleum refineries and chemical processes, MPC can now be found in a wide variety of application areas including chemicals, food processing, automotive, aerospace, metallurgy, and pulp and paper. A well-known implementation of MPC in chemical and refinery applications is Dynamic Matrix Control or DMC.
The MPC Controller employs a software model of the process to predict the effect of past changes of manipulated variable and measurable disturbances on the output variables of interest. The independent variables are computed so as to optimize future system behavior over a time interval known as the prediction horizon. In the general case any desired objective function can be used for the optimization. The system dynamics are described by an explicit process model, which can take, in principle, a number of different mathematical forms. Process input and output constraints are included directly in the problem formulation so that future constraint violations are anticipated and prevented.
In practice a number of different approaches have been developed and commercialized in implementing MPC Controllers. The most successful implementations have made use of a linear model for the plant dynamics. The linear model is developed in a first step by gathering data on the process by introducing test disturbances on the independent variables and measuring the effects of the disturbances on the dependent variables. This initial step is referred to as identification and the novel use of this identification data is the essence of this invention.
U.S. Pat. Nos. 4,349,869 and 4,616,308 describe an implementation of MPC control called Dynamic Matrix Control (DMC). These patents describe the MPC algorithms based on linear models of a plant and describe how process constraints are included in the problem formulation. Initial identification of the MPC controller using process data is also described.
By way of further background this Identification of process dynamics requires a pre-test in which the independent variables of the process are moved in some pattern to determine the effect on the dependent variables. In a chemical or refinery process the independent variables include the PID (proportional-integral-derivative) controller set points for selected dependent variables, the valve positions of PID controllers in manual, and temperatures, material flows, pressures and compositions that are determined outside the scope of the controller's domain. For any process Identification test, the independent variables are fixed for the analysis of the data. Further the tuning of any of the PID controllers in the domain of the MPC controller is fixed. The MPC controller that is built to use the dynamic process models from the Identification must have exactly the same configuration of independent variables that existed when the Identification was performed. Thus the PID controller configuration that is present during Identification imbeds the PID controller dynamics in the dynamic model.
This characteristic of current Identification technology represents an unsolved problem that is addressed by this invention. The problem creates a limitation on the use of MPC technology that manifests itself in two different areas.
The first application area is MPC itself. Because the dynamics of the PID controllers are imbedded in the MPC model, any change in the tuning of a PID controller or changing of the PID state from auto to manual or vice versa changes the dynamic model. To correct this it has been required to retest the process unit with the changed conditions. A well-designed Identification test for a complex multivariable process might be a 2-3 week effort involving careful planning and skilled people.
The second application area is in the field of Operator Training Simulators. Effective training simulators are important to the chemical process industry. The large investments in new chemical processes and the safety implications of the complex processes require a well-trained operator group. This is important especially for process units that remain on computer control for extended periods of time, since the operators do not have the opportunity to control the unit. MPC models are used in creating training simulators but MPC models obtained from current Identification technology fall short of ideal because of the aforementioned problem that the PID controller configuration that is present during Identification imbeds the PID controller dynamics in the dynamic model. The result of this is that authentic training is difficult because the operators cannot change the state (auto or manual) of the PID controllers without reducing the fidelity of the model. A survey of control rooms in the chemical process industry will reveal that they are rarely used after a start-up as the operating personnel learn that the simulator does not allow the operators to experiment with realistic control changes. A training simulator based on an Identification model that has the fidelity to hold a process at constraints, display all temperatures, pressures, flows, and valve positions and allow the operator to switch any PID controller to manual or auto would be a powerful tool for training.
Numerous unsuccessful attempts have been made by practitioners in the field to address this identification issue. One approach would be to run the identification test with the regulatory control scheme in manual. This of course fails because the process does not reach any type of steady state. Other attempts have been made to conduct a standard identification test with the regulatory scheme in place but to then set up the model with the valve positions as the independent variables. These approaches always lead to failure with erratic results. It has become recognized that this approach fails because the valve positions are correlated through the PID dynamics via measured and unmeasured disturbances that are always present in a real world identification test and are thus not independent.
The recognition of this fact and the method of removing the noise and unmeasured disturbances from the data set are the essence of the invention.